3D IIR-GRAPPA: a novel method for dynamic parallel MRI

Author(s)

Liu, K.; Zhang, J.; Yang, R.; Zhang, C.

Available versions

Abstract

Since the development of Generalized Autocalibrating Partially Parallel Acquisition (GRAPPA) [1], several algorithms have been proposed to incorporate it into the dynamic MRI applications. TGRAPPA tries to reduce the ACS acquisition by forming a composite frame via adding several neighbouring time frames, which are sampled in an interleaved pattern [2], and its image reconstruction of each frame is performed in a one-dimensional (1D) fashion by interpolating the linear correlation of the acquired data along phase encoding direction. Using the same composition and sampling scheme, k-t GRAPPA introduces a 2D reconstruction format, which interpolates the linear correlation of the acquired data in both phase-encoding and temporal directions [3]. The same as original GRAPPA [1], both these methods use moving average kernels with finite impulse response (FIR) for interpolation. We propose a novel 3D Infinite Impulse Response (IIR) temporal GRAPPA method. Different from [1]-[3], this method reconstructs the unacquired data points by interpolation along phase encoding, frequency encoding and temporal three directions, and uses autoregression (AR) moving average kernels with infinite impulse response in the interpolation. The AR part of the IIR kernel uses the previously reconstructed data points for further reconstructions, which captures more precisely the data acceleration along three directions and renders further reduction of required ACS lines and significant improvement of reconstructed images.

Publication year

2009

Publication type

Conference paper

Source

17th International Society for Magnetic Resonance in Medicine (ISMRM) Scientific Meeting and Exhibition, Honolulu, United States, 18-24 April 2009, p. 2733

Publisher

International Society for Magnetic Resonance in Medicine

Copyright

Copyright © 2009 The published version is reproduced with the permission of the publisher.

Details